Exercises for Multimedia Databases
Institut für Informationssysteme – TU Braunschweig - http://www.ifis.cs.tu-bs.de
Multimedia Databases
Exercise Sheet 07
(Audio Retrieval 2)
Please note: The exercises will be neither collected, nor corrected, or graded.
Exercise 1 – Query by Humming
a) How does a Query by Humming music recognition system work? Shortly explain the steps based on the architecture of such systems, presented in the lectures at slide 9.
b) How do we extract the melody from the wave file?
Exercise 2 – Parsons Code Representation
a) What is the Parsons code representation method?
b) How is matching with Parsons code performed and why can’t we do note to note matching?
c) What is the cost matrix and why do we need it?
d) Why should it be cheaper to replace R with U or D than replacing U with D or D with U?
e) On which assumption can we optimize the minimum cost path finding algorithm, and how do we do that?
Exercise 3 – Frame-based Representation
a) What are the advantages of frame based representation?
b) What is Dynamic Time Warping and why do we need it?
Exercise 4 – Hidden Markov Models
Considering the 5 elements of HMM, presented in the lecture in slides 71 and 72, draw a HMM (together with the ADSR representation of a note and corresponding observations), and explain each of these 5 elements.